Hacker News: loumaciel

New comment by loumaciel in "Show HN: LangAlpha – what if Claude Code was built for Wall Street?"

loumaciel — Wed, 15 Apr 2026 08:42:17 +0000

It becomes the only MCP. All other MCPs are registered within the tool.

New comment by loumaciel in "Show HN: LangAlpha – what if Claude Code was built for Wall Street?"

loumaciel — Tue, 14 Apr 2026 17:53:11 +0000

You can make MCP tools work for any type of data by using a proxy like https://github.com/lourencomaciel/sift-gateway/.

It saves the payloads into SQLite, maps them, and exposes tools for the model to run python against them. Works very well.

New comment by loumaciel in "Ask HN: What Are You Working On? (March 2026)"

loumaciel — Mon, 09 Mar 2026 13:23:52 +0000

I’m building Sift, a drop-in gateway that makes LLM tool use far more reliable when tools return large JSON payloads. The usual pattern is that agents paste raw tool outputs directly into the prompt, which quickly blows up context, causes truncation/compaction, and leads to incorrect answers once earlier results disappear. Sift sits between the model and its tools (MCP, APIs, CLIs), stores the full payload locally as an artifact (indexed in SQLite), and returns only a compact schema plus an artifact_id. When the model needs something from the data, it runs a tiny Python query against the stored artifact instead of reasoning over thousands of tokens of JSON. In benchmarks across 103 questions on real datasets, this approach cut input tokens by ~95% and improved answer accuracy from ~33% to ~99%. Repo: https://github.com/lourencomaciel/sift-gateway.

New comment by loumaciel in "Show HN: Keep large tool output out of LLM context: 3x accuracy 95% fewer tokens"

loumaciel — Thu, 05 Mar 2026 13:53:58 +0000

Happy to answer questions about the sandboxing, artifact format, or the benchmark setup.

The benchmark harness and datasets are in the repo if anyone wants to reproduce or extend the tests. Curious if others have run into the same context compaction issues with tool-heavy agents.

Show HN: Keep large tool output out of LLM context: 3x accuracy 95% fewer tokens

loumaciel — Thu, 05 Mar 2026 13:53:08 +0000

LLM agents often place raw JSON tool outputs directly in the prompt. After a few tool calls, earlier results get compacted or truncated and answers become incorrect or inconsistent.

I built Sift, a drop-in MCP gateway that stores tool outputs as local artifacts (filesystem blobs indexed in SQLite) and returns an `artifact_id` plus compact schema hints when responses are large or paginated.

Instead of reasoning over full JSON in the prompt, the model runs a small Python query:

    def run(data, schema, params):
        return max(data, key=lambda x: x["magnitude"])["place"]

Query code runs in a constrained subprocess (AST/import guards + timeout/memory caps). Only the computed result is returned to the model.

Benchmark (Claude Sonnet 4.6, 103 questions across 12 datasets):

- Baseline (raw JSON in prompt): 34/103 (33%), 10.7M input tokens

- Sift (artifact + code query): 102/103 (99%), 489K input tokens

Open benchmark + MIT code: https://github.com/lourencomaciel/sift-gateway

Install:

    pipx install sift-gateway
    sift-gateway init --from claude

Works with Claude Code, Cursor, Windsurf, Zed, and VS Code. Existing MCP servers and tools require no changes.

Comments URL: https://news.ycombinator.com/item?id=47261565

Points: 10

# Comments: 1

New comment by loumaciel in "US weighs Google break-up in landmark antitrust case"

loumaciel — Wed, 09 Oct 2024 19:37:07 +0000

What alternative promotional strategies?

New comment by loumaciel in "US weighs Google break-up in landmark antitrust case"

loumaciel — Wed, 09 Oct 2024 19:36:26 +0000

You don't rely on Google to handle everything, it's self-service. Whether you use Google or another platform depends entirely on where your audience is. For example, my father runs a small construction company, and 90% of his leads come from Google search. That’s where people are looking for services like his.

New comment by loumaciel in "US weighs Google break-up in landmark antitrust case"

loumaciel — Wed, 09 Oct 2024 19:32:21 +0000

Online advertising has leveled the playing field, allowing smaller brands to compete with big names. Platforms like Google make it easy to capture attention, which is why even giants like Nike are losing market share to newer players. This shift spans all non-regulated industries. Without online ads, launching a nationwide brand would require enormous budgets, leaving us stuck with the same old monopolies.

New comment by loumaciel in "US weighs Google break-up in landmark antitrust case"

loumaciel — Wed, 09 Oct 2024 19:03:06 +0000

Tons of small businesses rely on advertising. Tons of legit businesses only exist because online advertising became easy and widespread.

New comment by loumaciel in "Turning Everyday Gadgets into Bombs Is a Bad Idea"

loumaciel — Thu, 19 Sep 2024 23:01:00 +0000

In this analogy, what is water in the real world?

New comment by loumaciel in "Improvements to data analysis in ChatGPT"

loumaciel — Thu, 16 May 2024 23:14:51 +0000

https://www.findly.ai/ works with databases.

New comment by loumaciel in "Building ColdFusion for the Web"

loumaciel — Mon, 01 Jan 2024 00:36:01 +0000

Aha, I see, makes sense!

New comment by loumaciel in "Building ColdFusion for the Web"

loumaciel — Sun, 31 Dec 2023 15:06:53 +0000

Maybe they were expressing their confusion because you were speaking Spanish to people in Brazil?